This invention in general relates to interferometry and in particular to interferometric apparatus and methods by which the local surface characteristics of photolithographic stage mirrors or the like may be interferometrically measured in-situ to provide correction signals for enhanced distance measurement accuracy.
Interferometry is a well established metrology used extensively in microfabrication processes to measure and control a host of critical dimensions. It is especially important in manufacturing semiconductors and the like where requirements for precision are 10 to 40% better than critical dimensions of 0.1 xcexcm or less.
Integrated circuits made of semiconductor materials are constructed by successively depositing and patterning layers of different materials on a silicon wafer while it typically resides in a flat exposure plane having Cartesian x-y coordinates to which there is a normal z-direction. The patterning process consists of a combination of exposure and development of photoresist followed by etching and doping of the underlying layers followed by the deposition of subsequent layers. This process results in a complex and, on the scale of microns, very nonhomogeneous material structure on the wafer surface.
Typically each wafer contains multiple copies of the same pattern called xe2x80x9cfieldsxe2x80x9d arrayed on the wafer in a nominally rectilinear distribution known as the xe2x80x9cgrid.xe2x80x9d Often, but not always, each field corresponds to a single xe2x80x9cchip.xe2x80x9d
The exposure process consists of projecting the image of the next layer pattern onto (and into) the photoresist that has been spun onto the wafer. For an integrated circuit to function properly each successive projected image must be accurately matched to the patterns already on the wafer. The process of determining the position, orientation, and distortion of the patterns already on the wafer, and then placing them in the correct relation to the projected image, is termed xe2x80x9calignment.xe2x80x9d The actual outcome, i.e., how accurately each successive patterned layer is matched to the previous layers, is termed xe2x80x9coverlay.xe2x80x9d
In general, the alignment process requires both translational and rotational positioning of the wafer and/or the projected image as well as some distortion of the image to match the actual shape of the patterns already present. The fact that the wafer and the image need to be positioned correctly to get one pattern on top of the other is obvious. Actual distortion of the image is often needed as well. Other effects, such as thermal and vibration, may require compensation as well.
The net consequence of all this is that the shape of the first-level pattern printed on the wafer is not ideal and all subsequent patterns must, to the extent possible, be adjusted to fit the overall shape of the first-level printed pattern. Different exposure tools have different capabilities to account for these effects, but, in general, the distortions or shape variations that can be accounted for include x and y magnification and skew. These distortions, when combined with translation and rotation, make up the complete set of linear transformations in the plane.
Since the problem is to successively match the projected image to the patterns already on the wafer, and not simply to position the wafer itself, the exposure tool must effectively be able to detect or infer the relative position, orientation, and distortion of both the wafer patterns themselves and the projected image.
It is difficult to directly sense circuit patterns themselves, and therefore, alignment is accomplished by adding fiducial marks or xe2x80x9calignment marksxe2x80x9d to the circuit patterns. These alignment marks can be used to determine the reticle position, orientation, and distortion and/or the projected image position, orientation, and distortion. They can also be printed on the wafer along with the circuit pattern and hence can be used to determine the wafer pattern position, orientation, and distortion. Alignment marks generally consist of one or more clear or opaque lines on the reticle, which then become xe2x80x9ctrenchesxe2x80x9d or xe2x80x9cmesasxe2x80x9d when printed on the wafer. But more complex structures such as gratings, which are simply periodic arrays of trenches and/or mesas, and checkerboard patterns are also used. Alignment marks are usually located either along the edges of xe2x80x9ckerfxe2x80x9d of each field or a few xe2x80x9cmaster marksxe2x80x9d are distributed across the wafer. Although alignment marks are necessary, they are not part of the chip circuitry and therefore, from the chip maker""s point of view, they waste valuable wafer area or xe2x80x9creal estate.xe2x80x9d This drives alignment marks to be as small as possible, and they are often less than a few hundred micrometers on a side.
Alignment sensors are incorporated into the exposure tool to xe2x80x9cseexe2x80x9d alignment marks. Generally there are separate sensors for the wafer, the reticle, and/or the projected image itself. Depending on the overall alignment strategy, these sensors may be entirety separate systems or they may be effectively combined into a single sensor. For example, a sensor that can see the projected image directly would nominally be xe2x80x9cblindxe2x80x9d with respect to wafer marks and hence a separate wafer sensor is required. But a sensor that xe2x80x9clooksxe2x80x9d at the wafer through the reticle alignment marks themselves is essentially performing reticle and wafer alignment simultaneously and hence no separate reticle sensor is necessary. Note that in this case the positions of the alignment marks in the projected image are being inferred from the positions of the reticle alignment marks and a careful calibration of reticle to image positions must have been performed before the alignment step.
Furthermore, as implied above, essentially all exposure tools use sensors that detect the wafer alignment marks optically. That is, the sensors project light at one or more wavelengths onto the wafer and detect the scattering/diffraction from the alignment marks as a function of position in the wafer plane. Many types of alignment sensors are in common use and their optical configurations cover the full spectrum from simple microscopes to heterodyne grating interferometers. Also, since different sensor configurations operate better or worse on given wafer types, most exposure tools carry more than one sensor configuration to allow for good overlay on the widest possible range of wafer types.
The overall job of an alignment sensor is to determine the position of each of a given subset of all the alignment marks on a wafer in a coordinate system fixed with respect to the exposure tool. These position data are then used in either of two generic ways termed xe2x80x9cglobalxe2x80x9d and xe2x80x9cfield-by-fieldxe2x80x9d to perform alignment. In global alignment the marks in only a few fields are located by the alignment sensor(s) and the data are combined in a best-fit sense to determine the optimum alignment of all the fields on the wafer. In field-by-field alignment the data collected from a single field are used to align only that field. Global alignment is usually both faster, because not all the fields on the wafer are located, and less sensitive to noise, because it combines all the data together to find a best overall fit. But, since the results of the best fit are used in a feed-forward or dead reckoning approach, it does rely on the overall optomechanical stability of the exposure tool.
Alignment is generally implemented as a two-step process; that is, a fine alignment step with an accuracy of tens of nanometers follows an initial coarse alignment step with an accuracy of micrometers, and alignment requires positioning the wafer in all six degrees of freedom: three translation and three rotation. But adjusting the wafer so that it lies in the projected image plane, i.e., leveling and focusing the wafer, which involves one translational degree of freedom (motion along the optic axis, the z-axis or a parallel normal to the x-y wafer orientation) and two rotational degrees of freedom (orienting the plane of the wafer to be parallel to the projected image plane), is generally considered separate from alignment. Only in-plane translation (two degrees of freedom) and rotation about the projection optic axis (one degree of freedom) are commonly meant when referring to alignment. The reason for this separation in nomenclature is the difference in accuracy required. The accuracy required for in-plane translation and rotation generally needs to be on the order of several tens of nanometers or about 20 to 30% of the minimum feature size or critical dimension (CD) to be printed on the wafer. Current state-of-the-art CD values are on the order of several hundred nanometers, and thus, the required alignment accuracy is less than 100 nm. On the other hand, the accuracy required for out-of-plane translation and rotation is related to the total usable depth of focus of the exposure tool, which is generally close to the CD value. Thus, out-of-plane focusing and leveling the wafer require less accuracy than in-plane alignment. Also, the sensors for focusing and leveling are usually completely separate from the xe2x80x9calignment sensorsxe2x80x9d and focusing and leveling do not usually rely on patterns on the wafer. Only the wafer surface or its surrogate needs to be sensed. Nevertheless, this is still a substantial task requiring, among other things, precise knowledge about the vertical position (the altitude) of the optical projection system above the wafer.
To achieve alignment, it is known to use dynamic interferometers in which distance measurements are enhanced through the use of dynamic elements whose angular orientation is controlled via feedback arrangements to assure that beams carrying distance information are properly aligned to provide optimal signal. Such interferometers are shown, for example, in International Application No. PCT/US00/12097 filed May 5, 2000, and entitled xe2x80x9cInterferometry Systems Having a Dynamic Beam-Steering Assembly For Measuring Angle and Distancexe2x80x9d by Henry A. Hills Published Nov. 19, 2000 as International Publication No. WO 00/66969 and corresponding to U.S. Pat. No. 6,271,923 of the same title and inventor issued on Aug. 7, 2001. However, even with dynamic interferometers the shape of various reflecting elements impacts on the achievable accuracy in distance measurements and impacts on the achievable accuracy in angle measurements, because for the latter local slope changes influence beam directions, as stage mirrors undergo their various motions. Typically, the shape of such reflecting elements, such as thin high aspect ratio mirrors, is characterized off-stage and, if judged to be of adequate consistency, are then mounted on-stage. However, this is often unacceptable because the mounting process itself distorts the shape of the element compared with its inspected shape, and this change in shape can introduce measurement errors.
Accordingly, it is a major object of the present invention to provide interferometric apparatus and methods by which the shapes of on-stage reflecting elements, such as thin high aspect ratio mirrors, may be measured in-situ, after mounting, to develop correction signals that compensate for errors in optical path lengths and in beam directions related to shapes of reflecting surfaces.
It is another object of the present invention to provide interferometric apparatus and methods by which the shapes of on-stage reflecting elements, such as thin high aspect ratio mirrors, may be measured in-situ, after mounting, to develop correction signals that compensate for errors in optical path lengths and in beam directions related to shapes of reflecting surfaces arranged in orthogonal planes.
It is yet another object of the present invention to exploit information generated from the operating properties of dynamic interferometers by which the shapes of on-stage reflecting elements, such as thin high aspect ratio mirrors, may be measured in-situ, after mounting, to develop correction signals that compensate for errors in optical path lengths and in beam directions related to shapes of reflecting surfaces.
It is yet another object of the present invention to provide interferometric apparatus and methods by which the shapes of off-stage reflecting elements, such as thin high aspect ratio mirrors, may be measured in-situ, after mounting, to develop correction signals that compensate for errors in optical path lengths and in beam directions related to shapes of reflecting surfaces.
Other objects of the present invention will, in part, be obvious and will, in part, appear hereinafter when reading the following detailed description in conjunction with the drawings.
Interferometric apparatus and methods by which the local surface characteristics of photolithographic mirrors or the like may be interferometrically measured in-situ to provide correction signals for enhanced distance and angular measurement accuracy. Surface characterizations along one or multiple datum lines in one or more directions may be made by measuring the angular changes in beams reflected off the surfaces during scanning operations to determine local slope and then integrating the slope to arrive at surface topology. The mirrors may be mounted either on photolithographic stages or off the photolithographic stages on a reference frame. For the simplest case one dynamic beam-steering assembly or interferometer subsystem is employed for this purpose. For mirror characterization in two orthogonal directions, at least two dynamic beam-stearing assemblies are used. One produces a signal that contains information about the change in slope of the mirror surface along the datum line and orthogonal to it and the other produces a signal that contains information about the angular orientation of the stage on which the mirror is mounted. These two signals are combined to extract information about the slope of the mirror along its datum line and orthogonal to it. The slope is then integrated to obtain topography as a function of displacement. Single beam interferometers are preferred because they can measure pitch, yaw, and displacement with only a single beam to the mirror. Measurements can be made of a plurality of mirrors facing in mutually orthogonal directions by sequentially holding one or more fixed relative to their elongated surfaces while translating the third along its elongated dimension and repeating the process. Alternatively, all mirrors can be moved together to obtain relative mirror topography. Three beam-stearing assemblies may be used to fully characterize three corresponding mutually orthogonal mirrors and beam-stearing or interferometer subsystems may be mounted on or off the translation stage.
Once the mirror""s in-situ topography is established, it is stored in look-up-tables (LUTs), or the like, to provide real-time error correction signals to improve precision during normal operation.